+ All Categories
Home > Documents > Benefit of ASP, not generally being subjected to this:

Benefit of ASP, not generally being subjected to this:

Date post: 17-Dec-2015
Category:
Upload: diane-hortense-hood
View: 215 times
Download: 0 times
Share this document with a friend
41
Benefit of ASP, not generally being subjected to this:
Transcript

Benefit of ASP, not generally being subjected to this:

What the flux? What the flux? Constraining ecosystem models Constraining ecosystem models

with flux tower mesonetswith flux tower mesonets

Ankur DesaiNational Center for Atmospheric Research

ASP Research Review, 7 Mar 2007Boulder, CO USA

Carbon Dioxide• Carbon dioxide and climate are closely linked in

our atmospheric system

• Atmospheric mixing ratios of CO2 exceed anything seen in last 650,000 yr

Carbon Dioxide

Carbon Dioxide

Carbon Dioxide• Atmospheric CO2 growth rate

is not constant– more variable than rate of

increase in fossil fuel use

• Land and ocean sources/sinks– complex internal feedbacks– also affected by external

episodic (e.g., volcano) and oscillatory (e.g., ENSO) events

• Basic mechanisms understood– specific processes in land and

ocean are not– regional scale evaluation is

critically needed

Pools and Fluxes

The Terrestrial Ecosystem• Responses between land and atmospheric CO2 are

highly variable and functions of:– geography (e.g., N.H. land sink)– land cover– management (e.g., tropical deforestation)– land-atmosphere feedbacks of carbon, water and

energy

• Latest atmospheric data inversions and biogeochemical models converge on terrestrial carbon cycle as primary control on atmospheric CO2 growth rate variability (Peylin et al, 2005, GBC)

• Measurements of atmospheric CO2 over land have, until recently, been limited

Peylin et al, 2005, GBC

Terrestrial Ecosystem• Regional biosphere flux variability is complex

• Source: NOAA/ESRL (Carbon Tracker), units Mg Ha-1 yr-1

Terrestrial Terminology• The terrestrial CO2 cycle:

– Plants uptake CO2 by photosynthesis = Gross Primary Production (GPP) = function of light, CO2, water, temperature, humidity [Farquhar, Ball, Berry, Cook, Collatz, Sharkey]

– Plants respire some of this CO2 during carbohydarate conversion and utilization = Autotrophic Respiration (Ra) = function of temperature and substrate availability

– Soil bacteria decompose organic carbon (dead plants) and release CO2 back to the atmosphere = Heterotrophic Respiration (Rh) = function of temperature, soil moisture, substrate availability, bacterial community kinetics

– Total Ecosystem Respiration = Rh + Ra– Lots of non-linear interactions– Disturbance, land use, competitions are larger scale

effects

Terrestrial Terminology• Most important term:

– NEE = Net Ecosystem Exchange = Net CO2 flux = ER – GPP

• Negative = sink from atmosphere to biosphere• Positive = source from biosphere to atmosphere

• Modeling NEE, GPP, ER is hard because:– Functions are empirical, typically enzyme kinetics– Parameters are unknown, hard to measure– Works well for a single leaf, simple soil but not always for

entire forests and realistic soils• What are we trying to do

– Upscaling fluxes from leaf to forest stand, ecosystem, biome is current heart of research enterprise called the “bottom-up” approach

– Downscaling tracers/satellites from globe to continent to region is heart of the “top-down” approach

– Convergence = we can measure/predict/test hypotheses with regional fluxes

– At least 98 grad students agree and want to learn more

Measuring Stand Scale Flux• We can measure ecosystem land-atmosphere flux

(NEE) at spatial length scales of 1-10 km with the Eddy Covariance technique– How? Use the ensemble-averaged turbulent scalar

conservation equation

Measuring Stand Scale Flux• We have instruments to be able to do this

Measuring Stand Scale Flux

Respiration

Respiration and Photosynthesis

Measuring Stand Scale Flux

Measuring Stand Scale Flux• Top: Daily NEE, Bottom: Cumulative NEE

Measuring Stand Scale Flux

Measuring Stand Scale Flux• Lots of folks are now doing this (first in early 90s)

Pitfalls With Eddy Covariance• Major assumptions for using time-averaged flux

as stand-in for ensemble average (Reynolds’ “frozen field” hypothesis)– flow is turbulent, above roughness sublayer,

stationary– signal spectral attenuation and instrument lags are

minimal and can be empirically corrected– time period captures major scales of turbulence

Berger et al, 2001, JAOT

Pitfalls With Eddy Covariance• Nocturnal stable boundary layer provides most

challenging conditions:– nighttime NEE decline with u*

• suggests primary flow is not 1-D (e.g., advection)• intermittent turbulence

– non-homogenous cover/terrain effects

Cook et al, 2004, Ag. For. Met.

Desai et al, 2005, Ag. For. Met

Upscaling Goals• Upscaling fluxes from sites (e.g., measured with eddy

covarinace) to regions is a pressing research issue– Helps understand land-atmosphere interaction at scales

relevant to global models, decisions support– Emergent properties of land-atmosphere interaction may

appear– But: upscaling is hard when landcover or terrain is

complex• Hypotheses:

– Inversion of NEE from multiple tower sites can lead to regional scale ecosystem parameters that reproduce regional flux

– Parameters are significantly different across major ecosystem type boundaries

– Wetlands are more sensitive to precipitation variability than uplands

• Several regions have dense flux tower networks that could be used to constrain a regional ecosystem model

• Northern Wisconsin is one of these regions– Plus we can evaluate this flux with the 447-m tall flux

tower, tall tower ABL budgets, forest inventory, and a regional mesoscale CO2 inversion

Upscale This!

Already upscaled

Dense Mesonet

Tall Tower Cumulative NEE• Net annual source since 1997

Complex Landcover

Regional Flux?

Stand Scale Flux Variability

Method• We can use models constrained with data to get

regional flux• Ecosystem models do generally well at simulating

daily and seasonal cycle– Poor at interannual variability, long term trends– Also, parameters are unknown

• Parameter estimation using well established method – Markov Chain Monte Carlo (MCMC)

• Ecosystem Model to be used is SipNET• SipNET parameter estimation was designed from

the get-go to be “spatial”– Multiple sites can be assimilated at once– Some parameters vary spatially, others are fixed– Cost function reflects this by summing RMS model-

data error across sites and modifying parameter walk

Method• MCMC is an optimizing method to minimize model-data mismatch

– Quasi-random walk through parameter space (Metropolis)• Prior parameters distribution needed• Start at many places (random) in prior parameter space

– Move “downhill” to minima in model-data RMS– Avoid local minima by occasionally performing “uphill” moves– Requires ~100,000 model iterations– End result – “best” parameter set and confidence intervals

(from all the iterations)– NEE, Latent Heat Flux (LE) and Sensible Heat Flux (H) can all

be used• Nighttime NEE good measure of respiration, maybe H?• Daytime NEE, LE good measures of photosynthesis

• SipNET is fast (~100 ms year-1), so good for MCMC (hours)– Based on PNET ecosystem model– Tested at several sites– Driven by climate, parameters and initial carbon pools– Trivially parallelizable (needs to be done, though)

Simple Test of SipNET & MCMC

The Next Test• Region is 70% upland, 30% wetland• Combine the 3 hardwood sites together to

estimate upland NEE• Combine the 3 wetland sites to estimate wetland

NEE• Use remote sensing to add hardwood+wetland• Compare to using only 1 hardwood tower, 1

wetland tower, 1 hardwood+wetland tower• Compare to the independent regional flux

estimates (tall tower, FIA driven model, ABL budgets, regional inverse methods)

• See if parameters can predict interannual variability over next several years at tall tower

Progress• Not much, ACME07 and RBGC07 take all my time.

Need a catchy acronym to get more work done!• Test assimilation with tall tower done• SipNET probably not a good wetland model,

proposal funded to fix that• Number of parameters one can constrain with flux

data is relatively small (4-10), other data (transpiration, vegetation indices, …) could help– Meteorologists are better at this kind of data

assimilation but goal is different (forecast, equations are known, model is slower, [3,4]DVAR or EKF better suited)

• Could regional tracer mesonets also be used here?• Another oversampled test case this summer is the

North American Carbon Program (NACP) Mid-Continent Intensive (MCI) over Iowa

Conclusions• Atmospheric CO2 growth rates are mediated by land

fluxes– Problem is nonlinear - land fluxes are also functions of

CO2 and temperature• There’s lots to learn about land-atmosphere trace

gas exchange and interaction– Regional scales are key in terms of understanding

whole ecosystems, emergent responses, regional impacts, decision support and global model evaluation

• We can measure fluxes with the eddy covariance technique

• Scaling up and down is hard• Ecosystem models can be constrained with eddy

covariance flux data• Ecologists, meteorologists, foresters, and

hydrologists will one day live in perfect harmony

Thanks• Collaborators: Dave Schimel (CGD), Dave Moore

(CIRES), Steve Aulenbach (CGD), Ken Davis (PSU), Bill Sacks (UWI)

• Funding: NSF, DOE, NASA, USDA• Thanks: Land owners, technicians, students

Lots of Fluxes

WLEFtall tower

Lost Creekwetland

Willow Creekhardwood

Sylvaniaold-growth

Fluxes and Age

ABL Budget Equation


Recommended